It took exacycle 404,000 CPU-hours to do the full n2 comparison (it's an embarassingly parallel problem), but since Exacycle supplied over 600,000 CPU-hours per hour, it took less than one hour to compute. However, that wasn't "scalable" if you wanted to do full de novo assembly on every human (2.8e15 CPU hours) so Myers came up with a heuristic.
It took exacycle 404,000 CPU-hours to do the full n2 comparison (it's an embarassingly parallel problem), but since Exacycle supplied over 600,000 CPU-hours per hour, it took less than one hour to compute. However, that wasn't "scalable" if you wanted to do full de novo assembly on every human (2.8e15 CPU hours) so Myers came up with a heuristic.